Composition and Contrast: The Painterly Nature of Architectural Exterior Illumination

Abstract

CIE recommendations for architectural exterior illumination provide general guidelines for highlighting building forms, with emphasis on edges, curvature, and spatial layering. However, they do not explicitly address luminance contrast disposition—specifically, whether elements further from the viewer should appear brighter or if those closer should be more intensely lit. Inspiration for addressing this problem can be drawn from the principles of Renaissance and Baroque paintings, where techniques of working with light evolved from dramatic contrasts to more rational and balanced approaches, offering valuable models for contemporary illumination design. This study compares the principles of painting from that period with eye-tracking and survey-based methods to investigate whether the arrangement of luminance contrasts of illuminated building facades significantly influences viewers’ visual attention, aesthetic judgment, and perception of depth. The verification was conducted in two stages using three lighting variants of a selected architectural object. These variants differed in the luminance contrast distribution between surfaces closer to and farther from the observer, while maintaining a constant average luminance level across the entire façade of 10 cd/m². The first stage analysed visual reactions of 116 (out of 178) participants to luminance changes across the multi-segmented façade, presented in a darkened room on a luminance-calibrated display. The second stage involved a survey in which 358 participants were asked about their lighting preferences. Participants—including both design professionals and laypeople—exhibited consistent perceptions regarding how different lighting configurations affected their impression of the building. The results revealed that luminance disposition significantly influenced the perceived volume of the structure, particularly the sense of depth. Eye-tracking data also indicated a strong positive correlation between subjective aesthetic assessments and patterns of visual attention.

1. Introduction

In the context of rapidly developing urban environments, the exterior illumination of buildings is becoming increasingly significant (Zielinska-Dabkowska and Bobkowska 2022). This is not only due to functional reasons (Navarrete-Hernandez et al. 2025), but also economic (Giordano 2018), aesthetic (Żagan 2017), cultural (Cantizani-Oliva et al. 2024) and, more recently, environmental factors (Skarżyński and Żagan 2022). The character of architectural objects perceived in daylight can be significantly altered under artificial illumination. Light can both reveal the form of an object and give it an entirely new meaning (Balafoutis 2025). Except in special situations (Piekarski et al. 2025), however, it should not dominate the architecture but rather complement it, highlighting the most important elements of its form, emphasizing the architectural style of the building, its function, and integrating the object with its surroundings (Boyce 2019). In this sense, illumination becomes an art, a form of painting with light, where the light beam acts as the pigment. However, the design of illumination must take into account a range of technical and environmentally conscious aspects, as well as follow social expectations (Boyce 2019). This is especially important for historic buildings, where light as a new element enters a specific relationship with elements recognized as valuable (Ukabi and Akçay 2023).

Architectural illumination can be considered both in terms of successful implementations and unsuccessful ones that distort the appearance of buildings. In this context, norms and recommendations developed by international organizations operating in this field play a crucial role. Regarding architectural lighting from an aesthetic perspective, a particularly important document is the Guide for Floodlighting No. 94 (CIE 1993). Despite more than 30 years having passed, it remains relevant, and architectural lighting designers still follow its guidelines when illuminating buildings. International Commission on Illuminationin Report 94 (CIE 1993) provides detailed recommendations on luminance levels, methods of lighting buildings in the context of their tectonics and architectural style. It presents various approaches to lighting columns, horizontal and vertical division elements, and roofs. A particularly important piece of guidance for an architectural lighting designer is the schemes dedicated to buildings with different cross-sections. The report presents schemes for circular, square, and polygonal buildings, recommending optimal positioning of light fixtures and angles of incidence with respect to the observer’s location. The report also describes how to arrange lighting equipment for buildings with colonnades and entire structures. Following these recommendations aims to create lighting compositions that are visually attractive. However, the question arises: what does “visually attractive” at night actually mean (Gao and Zhu 2025) and how is it connected to visual behaviors?

In painting, there are several techniques referring to the realistic presentation of light and spatial relations on the canvas. The first is called aerial perspective (Șerbănescu 2024). It is a technique aimed at reproducing depth on the flat surface of the canvas by reflecting the natural changes that occur in the appearance of objects illuminated by natural light. Artists achieve the effect of depth by reducing the contrast and saturation of colours of the elements placed further in perspective. The farther the object is, the less intense its colours become. The hues also shift, becoming cooler as a result of light scattering in the atmosphere. Another effect is the brightening of distant objects, which enhances the impression of depth. With increasing distance, the details and contours of objects fade and eventually disappear. This technique is intended to imitate the way the human eye works and to convey the realism of the landscape during the day.

The second technique is tenebrism. Tenebrism is a painting technique characterized by the dominance of deep shadows and the use of a strongly contrasted, point-source light to illuminate selected elements of the composition. Emerging at the end of the sixteenth century, tenebrism created the effect of a dramatic “spotlight.” Its purpose was not only to emphasize the three-dimensionality of forms but also to strengthen the emotional impact and convey symbolic meanings (Xu 2020).

The character of contemporary urban lighting does not fully correspond to any of the classical principles of light manipulation known from painting. Light sources are multiplied, differing in colour temperature and direction of incidence. Such a situation may lead to distortions in the perception of distance and the size of objects, as well as significantly modify the character of space. Illumination, unlike natural light, which—regardless of weather conditions—has a uniform and comprehensive character, is often designed in a fragmentary manner, which creates additional interpretive difficulties.

Typically, colour temperature is not varied within a single object, and panoramas in which this aspect is consistently applied to emphasise depth are rarely encountered. Light sources are multiplied, differing in correlated colour temperature (CCT) and direction, which prevents full adherence to the principles used in Baroque nocturnal painting. Furthermore, in urban environments, for safety reasons (Liu et al. 2024)—it is difficult to apply strong contrasts based on areas entirely immersed in darkness. Such conditions may disrupt the perception of distance and object size, while also significantly altering the character of the space.

At this point, it is worth considering whether the principles of working with light in painting could be transferred to the technique of illuminating the urban landscape. This raises several research questions: Which principle should guide the design of architectural illumination? Should it be the principle of aerial perspective, treating illumination as a holistic influence on spatial perception, or rather the principle of tenebrism, emphasizing the directional and point-like character of light sources?

Can the painterly principles of contrast—developed in the practice of Baroque tenebrism—help to create an impression of depth in the illumination of architectural objects? If so, what rules should guide the design of lighting contrasts so that they do not disturb spatial perception but instead emphasize compositional and aesthetic qualities? How does a different distribution of light on a building’s surface affect the perception of its architectural character? Is it possible to consciously shape the perception of form through variations in the intensity and direction of illumination, a concept that directly relates to painting theory?

The CIE report Guide for Floodlighting No. 94 (CIE 1993) addresses the issue of contrast only in a brief paragraph:

“In nominally uniform lighting, the maximum: minimum ratio of 2:1 may be quite acceptable but a ratio of 3:1 may be obvious. No general limiting value of this ratio can be recommended because it is a matter of assessment and judgement on each site.”

Architectural lighting designers may have different objectives when choosing the distribution of contrasts, such as the desire to make the building visually attractive, adapting the lighting to the character of the building, or enhancing the legibility of the building’s form.

This raises a fundamental question: which planes should appear brighter? Those located deeper, as in daylight, or those closer to the observer, if the illumination is the opposite of daylight characteristics? Additionally, it should be verified whether the 2:1 contrast level indicated in the CIE report is indeed acceptable. Does it give the designer the flexibility to evoke distinct emotions and impressions, or does it become dominant, changing the order in which parts of the object are viewed?

The aim of the presented case study is to answer the research questions by combining survey research of non-professional observers with tracking the way they visually explore the illuminated building and familiarize themselves with its appearance. This approach is consistent with both the European Union’s policies on participatory design (European Commission 2019) and the guidelines for historic architectural sites (ICOMOS 2003).

2. Research Purposes and Hypothesis

The aim of this study is to analyse visual perception in the context of building façade lighting—particularly the influence of luminance contrast on the perception of architectural form at night. The main research question is: Does the use of the same luminance contrast—i.e., the difference between brightly lit and darker parts of the façade—but with different average luminance values assigned to surfaces located at different planes (closer and further away from the observer) affect the perception of the building’s depth? How does removing contrast between planes and unifying the average luminance across the entire building influence the perception of architecture?

The study adopted the guidelines of the CIE 95:1993 report as a reference point—both regarding the recommended contrast value and the average luminance level of the building as seen from a given viewing direction.

Subsequent questions focus on the perceptual effects of different luminance contrast variants—as well as their potential aesthetic and functional value. Essentially, the authors aim to examine whether observers as a group exhibit clear preferences for a specific luminance contrast variant—and whether there is a universal principle that makes one variant more visually appealing or compositionally legible than others.

In this context, three basic luminance distributions of the building should be considered:

A.

Variant where distant elements are lighter and closer ones are darker.

B.

Variant giving higher luminance to closer elements.

C.

Variant with an even luminance level across the entire building, eliminating contrast between planes to provide as uniform illumination as possible for all architectural elements.

The analysis of visual perception among a group of observers using eye tracking (ET) will also help answer the question of which contrast variant is more favourable from the perspective of perception and aesthetics. In the context of architectural lighting, “favourability” primarily refers to the effectiveness in attracting and maintaining the observers’ visual attention. These questions are significant because recent ET studies have revealed that higher luminance of an object observed at night does not always correlate with increased attention from observers (Kim et al. 2013). However, in cases where the shape and colour of an object cannot be modified, existing research and theories of perception can be utilized to consciously direct the attention of observers (Wagemans et al. 2012). The study will investigate whether changes in the distribution of luminance affect the perceived character of the examined architecture and the architectural form of the building (its dimensions).

In connection with the above, seven research hypotheses were formulated. The first four (Hypotheses 1–4) refer to visual behaviours recorded with the use of ET, while the remaining three (Hypotheses 5–7) concern the evaluation of the presented lighting solutions.

Hypothesis 1. 

The presence of additional contrasts on the façade located deeper in the courtyard will increase the number of fixations on the illustration and within the monument, while reducing the average fixation duration.

Hypothesis 2. 

Darker parts of an object will receive a shorter time of interest.

Hypothesis 3. 

Participants will spend less time looking at the façade of the monument when it is illuminated uniformly, without contrasts.

Hypothesis 4. 

A luminance contrast at the 2:1 level will not change the intuitive order in which the main parts of the monument are observed.

Hypothesis 5. 

There exists a lighting variant that will be preferred by the observers.

Hypothesis 6. 

A luminance contrast at the 2:1 level, as defined by the CIE report, is sufficient to evoke a sense of depth.

Hypothesis 7. 

Participants are expected to perceive a uniformly lit building without contrasts as calmer.

In addition, it is important to examine whether visual performance correlates with the way differently illuminated variants are evaluated. This will indicate whether, and to what extent, lighting designers can draw upon the experience and principles derived from painting theory.

A positive correlation between the data collected in the two parts of the experiment would mean that:

• The variant that participants observed for the longest time in the eye-tracking study will be the one rated highest in the survey.
• The variant identified as the least favourable will also correspond to the one with the shortest observation time.
• The variant perceived as the calmest will be the one that produced the most balanced distribution of visual attention.

In addition, due to the intention to evaluate the 2:1 luminance contrast, the correlation between Hypothesis 4 and Hypotheses 5–7 will be of particular importance. This relationship may provide the designer with the possibility of maintaining the natural order of viewing the monument’s façade while simultaneously evoking different perceptual impressions.

3. Materials and Methods

3.1. Methodology

The study will use ET technique, which involves recording and analysing human eye movements—it focuses on the points that a person looks at and the sequence and duration of fixations, meaning the time the gaze remains on a specific point (Duchowski 2007; Hutton 2019). Using ET we can evaluate cognitive processes (de Cock et al. 2022), identifying what attracts attention and what is ignored (Feng et al. 2022), how visual reactions are connected with architectural and urban elements, such as architectural details (e.g., facades), objects (e.g., columns, monuments), spaces (e.g., interiors, street panoramas) (Mandolesi et al. 2022; Mahmoud et al. 2022; Hamedani et al. 2020; Zvyagina 2018). ET allows objective and precise data collection, which until recently was only possible through subjective assessments, such as surveys. For such studies to be technically valid, several conditions must be met (Cassin 1990). Conducting objective ET research on illuminated objects in real conditions is difficult (Eghbal-Azar and Widlok 2013). The study assumes analysis of contrast, which requires changing the luminance of illuminated planes on the architectural object. Ideally, research would be conducted on a lit building with a lighting system that allows control of the luminous flux of the fixtures. However, this solution only makes sense for subjective studies, such as surveys, because uncontrolled external stimuli significantly affect the reliability of ET results. A possible solution is to build a full-scale mock-up. In dynamic scenarios, one can also use VR (Yi et al. 2024), which is theoretically possible but costly. Therefore, the authors decided to use virtual reality to simulate the lighting of the object. For this purpose, a computer simulation of the lighting of the complex of former government buildings at Bank Square in Warsaw was used (Krupiński 2024). The central part of the complex features depth (Figure 1), making it a very suitable base for scientific research.

The simulation presented in Figure 1 is highly photorealistic. However, it contains numerous stimuli that could attract the observer’s gaze and affect the study results. These include cars, the Moon, and many small-area contrasts. To address this, the luminous flux of the lighting fixtures was modified to equalize luminance levels only on elements that are significantly displaced relative to each other. At the same time, the minimal acceptable contrast recommended by the Guide for Floodlighting (2:1) was applied. The contrast is presented both as positive (Figure 2a) and negative (Figure 2c) in relation to the atmospheric perspective used in painting. To verify the role of contrast in architectural illumination, the luminance across the entire frame visible to the human eye was equalized (Figure 2e). The average luminance in the field of view was set at 10 cd/m², in line with the CIE report recommendations. Each simulation was monitored using FalseColour images (Figure 2b,d,f). The study monitor was placed in a dark room without external distractions. Prior to the study, the monitor was luminance-calibrated using a matrix luminance meter from TechnoTeam. The calibration of a screen was performed in a photometric dark room. The consistency between the simulation and the image displayed on the monitor was verified using a matrix luminance meter.

3.2. Participants

It was decided that only people with no expertise in the field under question will be invited to participate in the experiment. The way professionals look at an object belonging to their area of interest (AOI) differs considerably from that of lay people (Hekkert and van Wieringen 1996; Fudali-Czyż et al. 2018). The participants were financially remunerated. The authors had wanted the participantsnot to be familiar with the object they were looking at; therefore, the eye-tracking registration was not done in Warsaw. Finally, 178 adult citizens of Wrocław, Poland, took part in the study.

Having been briefed about the purpose and the process of the experiment, the participants signed a consent form to take part in the study.

Participants (or the visual data collected from them) were disqualified if:

-

Their eyesight was significantly impaired, and they did not wish to correct it in any way. Participants were expected to see correctly all aspects of the images shown at approximately 60 cm from their eyes. An examination done by an optometrist was to exclude monocularity, strabismus, and problems with perceiving color (Choudhury 2014) and contrast (Howell movement cards and tests including Worth, MARS, Ishihara, and Farnsworth D15 tests).

-

calibration parameters were not deemed satisfactory (max. error < 0.50° and the av. error < 0.30° (Carter and Luke 2020)

-

they failed to follow the procedure (talked or moved, ringing or vibrating phone while looking at the image);

-

they recognized the displayed object, (name of the object or the name of the city in which the object is located).

As a result, visual data collected from 116 participants was finally used in the analysis. The participants were all aged 19–60 (average = 30 years old, median = 25 years old, standard deviation = 9 years), and 55 were men and 61 were women. Out of these people 43 were shown image A, 36 were shown image B, and 37 were exposed to image C. Grouping was a crucial aspect of the experimental design, as presenting all three variants to each participant would have introduced confounding factors. Participants, having prior exposure to a specific version, may have altered their viewing behavior when examining subsequent examples. This could lead to potential biases, such as increased scrutiny of areas undergoing changes or a sense of boredom. It is important to note that the composition of the groups encompassed both male and female participants, aligning with the demographics observed in the 2021 study results (Hood and Walther 2021). Deviations in the aforementioned factor were not identified as a potential influencing element on the outcomes of the study.

3.3. Procedure

The first experiment was controlled using Tobii Glasses 2 and TobiiProGlasses software. The first step of the procedure was one-point calibration. It was conducted in a dark room, where the eyes had to adapt to the conditions of reduced luminance. If the software failed to calibrate correctly, it was repeated up to five times, provided that the search for the reason for the failure took no longer than five minutes. If calibration went smoothly, the recording that followed took around six and a half minutes. The entire experiment consisted of several slides with instructions and images. It was important that the participants do not make use of their short-term memory so three different sets of images were prepared, each including just one of the stimuli showing the palace with a different type of illumination. The image, be it variant A, B or C, was always displayed as image number 22 in the sequence of all 27 images. Each image was visible on the screen for 10 s. Before each image, the participants saw the question “Do you know this place?”, which would result in a sustained and known cognitive intention on the part of the participants (Yarbus 1968). Extending the duration of the experiment with additional stimuli was important; its purpose was to allow the observers to adapt to staying in the darkroom and to become accustomed to participating in the study. Before seeing the test illustration, the participants remained in a room illuminated only by the monitor light for 8–10 min. During the survey carried out after the eye-tracking test it was checked whether the participant recognized the shown building or not.

Data on observer preferences were collected from a separate group of 352 participants throughan online questionnaire. The division was intended to eliminate the potential influence of a previously viewed variant on participants’ preferences. The participants looked at visualizations A, B, and C, as well as two other visualizations of the same object featuring additional contrasts. The additional variants were intended to allow the choice of another preferred option that was completely unrelated to the aim of the study.All adult participantswho declared that they were completing the survey in a dark room were eligible to participate. All the tested examples were visible on the screen. Besides providing information on gender, age, profession, visual condition, and the type of screen on which they viewed the examples, participants were asked to indicate:

• the variant they liked the most and the one they liked the least.
• the variant that best emphasizes the official character of the building.
• the variant that gives the strongest sense of calm.
• the variant that makes the building appear wider.
• the variant that creates the strongest sense of spatial depth.

4. Results

The analysis of the data collected during the ET study was divided into three stages. Stage I involved the segregation and systematization of the research material. Stage II consisted of a general analysis of the observers’ visual engagement with the entire visualization and the monument as a whole (AOI 1 and AOI 2) considering the different illumination variants. Stage III focused on a more detailed analysis of the attractiveness of individual planes, concentrating on the successive parts of the building (AOI L, R, M). In separate stage the subjective preferences of the observers, collected through a questionnaire, were examined.

4.1. Visual Data Segregation and Pre-Processing

Data were generated and processed in Tobii Pro Lab (v. 1.194) software. Recordings in which the registered process of gaze-shifting lasted less than 85% of the entire time of the experiment were not included in the analysis. Those deemed credible were used as basis for assisted mapping and manual quality control of projection of fixation points. Then the areas of visual interest were established (Figure 3). AOI M, AOI L and AOI R corresponded to areas whose luminance changed between the stimuli. After all this was done, numerical reports in Tobii Pro Lab were generated. The reports (.xlsx) included the following parameters: Number of Viewers, Total Visit Duration (TVD), Average Fixation Duration (AFD), Fixation Count (FC), Time to First Fixation (TTFF). The data was put in order; charts were generated and the significance of deviations between the three variants was analysed statistically using Statistica 3.11 and DATAtab software (2025).

4.2. Data Analysis

4.2.1. Eye Tracking Data

One parameter that did not show significant deviation between stimuli A, B, and C was the Total Visit Duration within AOI 1. However, a simple box-plot comparative analysis (Figure 4, graph on the left) shows that stimulus A was the most visually engaging. The dark centre in stimulus B made the entire image less attractive visually, which is manifested in the fact that the volunteers exposed to this image spent the most time looking outside it (0.5 s on average, highest value equalled 1.15 s). This indicates the lower attractiveness of Variant B.

The first parameter that exhibited significant deviation between the three stimuli was the Total Visit Duration within AOI 2—which means the total amount of time spent looking at the monument in its entirety. The analysis of data presented on a chart (Figure 4 graph on the right) shows that those exposed to stimulus A spent the longest looking at the monument. Stimuli B and C were in that respect much less engaging. The dark centre of image B resulted in participants spending considerably more time looking outside the monument—including at the dark sky, devoid of any details. The brighter centre part of the building attracted attention, the darker centre caused the observer to avoid looking at it, and the uniform illumination was perceived as neutral. This does not confirm Hypothesis 3, as the lack of contrast turned out to be more visually attractive than the presence of contrast in Variant B.

The next part is related to the analysis of the entire image (AOI 1) as well as to the perception of the entire monument (AOI 2). It was concluded that the changes in illumination did not have a significant effect on how visually engaging both AOI 1 and AOI 2 were. This refers to Fixation Count as well as to the Average Fixation Duration. On average FC for variants A, B, and C amounted to 24–26 fixations, and the AFD within AOI 1 equalled 300–350 ms for all three stimuli. An ANOVA test (Kruskal-Waliss or one-way) was used for all parameters mentioned in this paragraph, where p equaled more than 0.05.

The analysis of the three distinct parts of the building (AOI L/AOI M/AOI R—Figure 3) showed that almost every participant of the experiment looked at least once at each of the areas. In addition, the analysis of VC proved that all volunteers shifted gaze between different AOI a similar number of times for all variants. Most participants (55.2%) visited AOI M five or six times. On average they also shifted attention between AOI L and AOI R 5–6 times. It was checked whether different types of illumination caused deviations in the preferred order of looking at different parts of the monument. This was diagnosed by comparing the average time that passed before the participants looked at the given AOI (TTFF). For AOI M this amounted to a value between 0.01 s and 0.05 s (ANOVA p < 0.05). 100% of participants, no matter which of the three stimuli they were exposed to, looked first at AOI M. Then their attention was usually shifted to AOI L (75% of all participants did that—69%, 75%, and 81% of the people for groups shown stimuli A, B, and C respectively). AOI R tended to be the last part to be taken in. This makes further analysis possible, since one knows now that the cognitive situation for all three groups was similar to a large degree. This constitutes an interesting additional observation, suggesting that despite their differences, the prepared lighting variants did not distort the overall hierarchy in the way observers viewed the monument. This result confirms Hypothesis 4.

The analysis of the TVD for central part of the building (Figure 5, AOI M) showed distinct differences across the tested lighting scenarios. The ANOVA test confirmed statistically significant differences between the scenarios (p = 0.00), indicating that the luminance level of courtyard illumination had a substantial effect on the allocation of visual attention to the central area of interest. The longest average TVD was observed in visualization A (with a brighter centre of the courtyard), reaching 4.64 s. In contrast, the shortest average duration of fixations on the central composition occurred in stimuli B, where the courtyard lighting exhibited the lowest luminance (3.55 s). Example C represented an intermediate condition with an average of 3.71 s. This result provides empirical support for Hypothesis 2. However, similar deviations were not observed with respect to the way participants maintained visual attention (TVD) on the wings of the monument (Figure 5, AOI L + R). Although statistically significant differences were indicated by the ANOVA test (p < 0.05), the viewing patterns for the wings in Examples A and B (2.62 and 2.70 s) were essentially the same, while the longest observation time for the side colonnades occurred in variant C. Therefore, this part of the analysis does not allow Hypothesis 2 to be fully confirmed.

4.2.2. Survey Data

The online survey was completed by 358 participants, including 215 respondents identified as experts (architects, conservators, individuals with electrical or lighting engineering backgrounds, etc.) and 143 non-professionals. Nine responses were excluded from further analysis due to respondents indicating impaired vision at the time of the survey, such as wearing dirty or damaged glasses or experiencing difficulties with night vision. One additional response was eliminated because of internal inconsistency. The initial comparative analysis of the responses of experts and non-professionals clearly showed no significant differences in the responses regarding all diagnosed aspects. Therefore, the responses were presented together. The analysis included 170 women, 173 men, and 5 respondents identifying with other genders. The average age of the verified respondents was 40 years, with the youngest being 18, the oldest 81, and the median age also 40. The answers (

Table 1. Survey results—respondents’ answers according to their subjective assessment. Left column: questions asked to participants. Top row: possible answers.

	All		A	B	C	Hard to Say	None/Does Not Affect Me/Other
Which variant do you like the most? (single selection)			235	10	36	4	63
Which variant do you like the least? (single selection)			23	219	27	40	39
Which variant gives you the strongest sense of calm? (single selection)	13		18	20	250	47
Which variant best conveys the building’s official character? (multiple selection)			163	8	78	14	110
Which variant makes the building appear wider? (multiple selection)			27	112	107	32	34
Which variant creates the strongest sense of spatial depth? (multiple selection)			35	198	8	5	102

) were collected as presented below:

The survey results proved to be unambiguous—the observers, as a group, demonstrated clear preferences regarding the distribution and arrangement of luminance on the presented façade. A total of 235 respondents (68%) indicated that example A was the most appealing. Only 10 respondents (3%) selected example B. Variant C was chosen as the most favourable by 36 respondents (10%), while 4 participants (1%) were unable to decide and 18% decided to choose two other variants (D-12% and E-6%). Variant B was indicated 219 times (63%) as the least liked option. The remaining examples received between 23 and 39 negative responses. The differences in response frequencies concerning preferences are statistically significant according to pairwise proportion z-tests. Specifically, option A was selected by a much larger proportion of respondents compared to all other options. Even after Bonferroni correction for multiple comparisons, the dominance of option A remains highly significant (p < 0.001). Hypothesis 5 was clearly confirmed.

The survey included a question concerning the impression of spatial depth. The subjective results confirmed that luminance contrast 2:1 influences the perception of depth, confirming Hypothesis 6. 198 respondents perceived the courtyard as deeper in variant B. Pairwise proportion z-tests revealed statistically significant differences between response categories. Option B was selected by a much larger share of respondents than all other options, and this result remained highly significant after Bonferroni correction (p < 0.001). In contrast, options C and “Hard to say” were chosen very rarely, showing significantly lower proportions compared to both A and B.

Respondents evaluating which variants visually widened the building showed no clear overall preference. Pairwise proportion z-tests indicated that options B (35.9%) and C (34.3%) were chosen most frequently, with no significant difference between them. Option A (8.7%) was selected significantly less often than both B and C (p < 0.001 after Bonferroni correction). The categories “Don’t know” and “Other” represented intermediate proportions and did not differ significantly from A, B, or C.

The survey included a question on which lighting variant best reflects the official character of the building. Pairwise proportion z-tests showed that option A, indicated by A 163 respondents (43.7%), was chosen most frequently and significantly more often than all other categories (p < 0.001 after Bonferroni correction). Option “Other” (29.5%) ranked second, followed by option C (20.9%), both of which were selected significantly more often than options B (2.1%) and “Don’t know” (3.8%). Thus, option A clearly dominated the responses, which strongly confirms Hypothesis 7.

5. Discussion

The aim of this article was to address three key questions related to the design of architectural lighting from an aesthetic perspective: which surfaces should be brighter—those closer or further away; whether a luminance contrast of 2:1, as recommended by the CIE 94:1993 report, is sufficient in terms of visual perception and attractiveness; and whether a combination of computer visualization techniques, eye-tracking, and surveys can contribute to the development of a contemporary sustainable theory of illumination.

The results show that Hypotheses 1–3 were not confirmed, while Hypotheses 4–7 were confirmed. Importantly, the expected correlations between eye-tracking data and questionnaire responses did occur. Variant A, which participants observed for the longest time, was also rated the highest in the survey. Variant B, associated with the shortest observation time, was identified as the least favourable. Finally, Variant C, perceived as the calmest, produced the most balanced distribution of visual attention.

The variant inspired by the principles of aerial perspective (Variant A) was effective in capturing attention and enhancing attractiveness; however, the absence of cooler hues and reduced contrasts for distant planes means that it does not convincingly convey depth. In contrast, the Baroque-inspired variant with tenebrism (Variant B) was found to be least attractive in the eye-tracking study, although respondents perceived the darker façade as more distant. This indicates that a simplified application of the theory of aerial perspective in illumination design may be misleading. The findings suggest that the appeal of illumination does not result from maximizing contrast or brightness, but rather from the appropriate distribution of luminance, the interplay of light with architectural form. According to the observers’ impressions, both experts and non-professionals, this does not necessarily correspond to the enhancement of spatial perception.

Although the CIE 94 report does not impose strict restrictions on increasing contrast, the minimum value of 2:1 appears sufficient in light of this research. Lighting designers should avoid uncontrolled multiplication of contrasts, as this may lead to excessive and chaotic illumination. Instead, the focus should be placed on thoughtful composition, harmony with the architectural character, and moderation. Proper management of contrasts—rather than their maximization—remains the key to visually attractive illumination, just as in the principles of painting.

Future studies should verify other luminance contrasts, such as 3:1, and extend the number of tested scenarios. It is also necessary to examine illumination across different scales: at the detail level (colonnades, arcades), at the urban scale (river or seaside panoramas, terraced buildings on a hillside), and in different contextual settings such as streets, squares, or natural surroundings. Further research should address the role of colour temperature modification and selective contrast reduction or enhancement in shaping the perception of depth. Surveys remain subject to technical limitations (e.g., screen brightness beyond the researchers’ control) but combining them with objective eye-tracking provides valuable complementary insights.

Possibly that the interplay of light and architectural form is inherently complex and cannot be fully reduced to simple rules of brightness and contrast. However, research of this kind may serve as the foundation for developing efficient strategies to optimize illumination in practice. Such strategies could take the form of design guidelines, interactive toolkits, or software applications that assist lighting designers in making context-sensitive decisions. In this process, the use of artificial intelligence will be indispensable, enabling the integration of large datasets from ET experiments, survey responses, and computational modelling to generate predictive insights. AI-driven applications could help simulate how observers are likely to perceive different lighting scenarios, allowing designers to optimize aesthetic quality, energy efficiency, and cultural sensitivity before implementation.

This study confirms that combining objective methods (eye-tracking) with subjective evaluations (surveys) is a valid approach to developing a sustainable theory of illumination. This extends previous research on sustainable and aesthetically appealing ways of presenting nightscapes (Gao and Zhu 2025). At the same time, it raises further methodological questions: should advance of tools such as eye-tracking, artificial intelligence algorithms, or 3D modelling of visual perception become standard practice in lighting design? Answering these questions will be crucial for the future of responsible, energy-conscious, and aesthetically appropriate architectural illumination.